April 24, 2008
Chapter 3
Theory and Methodology
1.Introduction
1.1.Overview
1.2.A review of the methodological points in chapter 2
Repeatable phenomena
*Schemas and okSchemas
*Examples and okExamples
1.3.Outline of the chapter
2.Theory of judgment making
2.1.Introduction[1]
The main goal of generative grammaris understood to be the discovery of the properties of the Computational System (CS) whose input is a set of items taken from the mental Lexicon of the speaker and whose output is a pair of mental representations – one underlying 'meaning' and the other 'sounds'.
(1)Numeration ==> CS ==>LF and PF
Once we take informant judgments to be a primary source of data for theory construction and evaluation, we must recognize the interrelation between a theory of the CSand a theory of how the informant's judgment obtains. The former cannot be empirically evaluated without the latter; considerations about the latter, on the other hand, could be revealing about the properties of the CS only if the former makes crucial reference to the latter.
(2)a.a theory of the CS
b.a theory of how the informant's judgment obtains
(3)a.(2a) cannot be empirically evaluated without (2b).
b.(2b) is useful to an empirical evaluation of (2a) only if (2b) makes crucial reference to (2a).
The relation between (2a) and (2b) can be established if (2a) is embedded within (2b). One way to embed (2a) within (2b) is to assume that the informant, upon seeing or listening to the presented sentence, forms a numeration, and compares the output of the CS with the presented sentence itself along with the interpretation being considered. In the absence of alternative ways of embedding (2a) within (2b), we will in fact pursue this possibility, following the recent works by A. Ueyama.
Let us consider how the informants make their judgment on sentence under interpretation , focusing on experiments where questions such as those in (4)are part of the instructions to the informants.[2]
(4)a.Is sentence acceptable under interpretation ?
b.How acceptable is sentence under interpretation ?
'Interpretation' in (4) is [thus<= I didn’t get the logical connection here.] not meant to be the interpretation of entire it is instead meant to be part of the interpretation of sentence . More in particular, what is intended by 'interpretation ' in (4) is an interpretation involving two linguistic expressions such as an anaphoric relation holding between two expressions. A more accurate way of stating the question in (4a) might therefore be something like "Is sentence acceptable under an interpretation that includes ?" I will however avoid using such a cumbersome way of phrasing the question and continue to use 'under interpretation ' in the ensuing discussion unless a more precise formulation is crucially needed.
In section2.2, I will introduce Ueyama's model of judgment making by the informant. It will be argued that we are led to that model as the result of adopting the model of CS as in (1) and accepting the thesis that informant judgments constitute a primary source of data for construction and evaluation of our theories about the CS, [combined with the desire to make generative grammar an empirical science.<= At first it wasn’t clear which part this modifies. I understand that it modifies "the thesis". Is that right?] The CS being embedded in a performance model poses inherent difficulty in accessing data about the properties of the CS, and we will? address the difficulty in section 2.2.2.
In section 2.3, a particular conception of the sense of acceptability will be introduced as a consequence of the considerations in section 2.2, [reflecting the difficulty just alluded to.<= What does this modify?] In section 2.4, we will recognize, in relation to the reliability of data about the properties of the CS, that the informant judgments are revealing about the properties of the CS only if the informants pay close enough attention to sentence and interpretation . We will further recognizein that section that the informants'resourcefulness may affect their judgments.
2.2.The model of judgment making and the CS
A more articulated version of what is suggested in chapter 2.
2.2.1.How the theory of the CS is embedded in the model of judgment making
[KK]このsectionはいつも読むのが辛く、頭を使います。私自身がきちんとした見方をまだ持てていないことにも拠ると思います。でも、今回のversionはかなりまとまって、読みやすくなったとは思いますが。If you find it読むのが辛i, I think most other readers will find it the same way. So, I think I should work on this section to make it easier to follow.
Consider again the model of the CS in (1), repeated here.
(1)The Model of the CS:
Numeration == CS ==>LF and PF
A theory of the CS is concerned with the properties of the CS,: what numeration gets mapped to what LF-PF output, what operations combine the items in the numeration in the derivation, what restrictions are imposed, if any, on such operations, etc. As noted before, neither a numeration nor a LF-PF output of the CS is directly accessible to us. What is accessible to us is the informants' reported judgments on the sentences presented to them.
(5)Presented Sentence ≈≈> ??? ≈≈> Judgment
Insofar as we take informant judgments to be a primary source of data for constructing and evaluating hypotheses about the CS, it seems natural, if not unavoidable, that ??? in (5) iscontains?Numeration == CS ==>LF and PF in (1).
In other words, if the CS plays the central role when the informant makes her judgment on an example sentence, it seems natural to assume that the CS is embedded in the model of judgment making by the informant, as schematized in (6).[3]
(6)Presented Sentence ≈≈> Numeration ==> CS ==>LF and PF ≈≈> Judgment
Recall that if we did not adopt this view, it would not be clear how the model of CS is related to the judgment making by the informant, which in turn would make it quite unclear in what sense informant judgments serve as data for research concerned with the properties of the CS.
According to this view, which has been advocated in works by A. Ueyama, the informant forms a numeration in part on the basis of the presented sentence, along with the 'intended interpretation' as provided in the instructions, and tries to determine whether the output of the CS on the basis of that numeration is compatible with (i) the phonetic string of the presented sentence and (ii) the 'intended interpretation'. [If we assume that the informant would always form a numeration that would end up yielding the PF that is compatible with the presented sentence, we can be concerned only with thate aspect of the 'compatibility check' by the informant that which has to do with the 'intended interpretation', rather than the compatibility between the presented sentence and the PF output.[4]<= I had difficulty understanding this sentence. I made some changes so that the modification relations would be clearer. I’m not sure if you want this, though.]
Something must mediate the presented sentence and a numeration. [I assume that? Wwhat mediates the two will be called Parser, following Ueyama to appear (2007 JK).<= I made a change to avoid a dangling participle (i.e., the "following...” part).] We take the sentence presented to the informant to be the input to Parser – which, in part on the basis of word recognition, identifies the argument-predicate relations and the modification relations among the elements in the sentence – and take the output of Parser as contributing to the formation of a particular numeration.[5] We can thus modify (6) as in (7).
(7)Presented Sentence ==> Parser ≈≈> Numeration == CS ==>LF and PF ≈≈> Judgment
Assuming that the LF representation gets mapped to a semantic representation (SR) (by the mapping of LF syntactic objects to semantic objects), and suppressing the PF output of the CS in our present discussion (see below), the model of judgment making by the informant can be schematized as in (8).
(8)Presented Sentence ==> Parser ≈≈> Numeration =input= CS =output=LF ==> SR ≈≈> Judgment
As noted above, what is intended within the box in (8) is that the input to the CS is a numeration and that? LF is an output of the CS. The two arrows before and after CS in (7) thus represent the 'is the input of' relation and the 'yields as an output of' relation, respectively, as indicated in (8). Similarly, what is meant by the arrow between LF and SR is that LF is the input to the mapping operations and SR is the output. A numeration thus necessarily results in SR(). The other instances of the arrows in (8) are used more loosely, as indicated in (9).
(9)a.Presented Sentence ==> Parser: ... is part of the input to ...
b.Parser ≈≈> Numeration: ... contributes to the formation of ...
c.SR ≈≈> Judgment: ... serves as a basis for ...
Recall that according to the model of the CS adopted here, its input is a numeration (i.e., a set of items taken from the Lexicon, possibly along with some features) and its two output representations are LF and PF (representations).
(1)Numeration ==> CS ==>LF and PF
(10)The CS:[6]
Input= Numeration
Output= LF() & PF()
As noted before, it is necessary to embed the CS in the model of judgment making once we accept that the informant judgments are a primary source of data for construction and evaluation of the theory of the CS. Once the CS is assumed to be embedded within the model of judgment making, the presented sentence, along with the 'intended interpretation', must help the formation of the numeration, as also noted above. Hence the conception of Parser adopted here is also a consequence of adopting the basic assumptions adopted here.
Let us review the model of judgment making by the informant. The informant is presented with sentence and is asked a question such as (4a), repeated here.
(4)a.Is sentence acceptable under interpretation ?
For example, the question can be "Is John thinks Mary loves himself acceptable under the interpretation that John and himself refer to the same individual.[7] The presented sentence is part of the input to Parser since interpretation as well as the informant's world knowledge might also help Parser identify the argument-predicate relations and the modification relations among the elements in the sentence; see (9a).[8] On the basis of such information, Parser helps the formation of a numeration; see (9b).[9] Once numeration has been determined, the CS yields two output representations LF() & PF() by the application of some mechanical operations. LF() gets mapped to a semantic representation SR() by transparent and mechanical application of some mapping rules that convert LF objects to SR objects. Assuming that PF() is non-distinct from , the informant's answer to (4a) is based on whether SR() is compatible with , i.e., it is about the compatibility between SR() and . In that sense, the intended interpretation as well as SR() serve as a basis for the judgment; see (9c).
2.2.2.Difficulty in 'accessing' data about the CS
2.2.2.1.Introduction
When presented a question such as (4a), the informant tries to come up with numeration primarily on the basis of and (see above) such that would serve as the input to the CS. As noted, we assume that as long as the informant pays close enough attention to the form of in judging it – i.e., as long as the informant is indeed judging the presented sentence – the PF(), if it obtains, necessarily corresponds to Given that LF is a basis of 'meaning', it is reasonable to assume that the informant obtains the 'meaning' of on the basis of LF(), and such a 'meaning'is referred to here as SR(). The informant then determines whether the SR()is compatible with in (4a). If it is, the answer to (4a) is yes; and otherwise, the answer is no. The informant's sense of acceptability on sentence under interpretation is thus a function of (11).
(11)whether or not the informant can 'come up with' numeration on the basis of such that SR()compatible with obtains.
Recall that that numeration and SR() are an input to the CS and an output of the CS, respectively, i.e., the relation between and SR() is determined mechanically by the CS.[10] Information regarding the relations between numerations and SRs would therefore contribute to our understanding of the properties of the CS.
There are, however, two obstacles we must deal with in obtaining information about the relations between numeration and SR(). They can be stated as in (12).
(12)a.We cannot tell what numeration the informant 'goes to' upon seeing or listening to the presented string.
b.We cannot rule out the possibility that the informant's reported judgmenton the acceptability of sentence under interpretation has been[unaffected<=?] by the possible unnaturalness of the entire SR().
(12a) has to do with Parsing hence before the box in (8) while (12b) is concerned with after the box in (8). In this subsection, we will briefly address the two issues.
2.2.2.2.Difficulty in Parsing
The informant might encounter difficulty in 'going to' the 'right' numeration corresponding to that would result in SR() compatible with . The difficulty may be due to the complexity of the entire sentence . But it can also be due to the ambiguity of ; i.e., the surface string may correspond to more than one numeration and hence more than one LF representation. In this subsection, I provide a brief and schematic illustration of a couple of specific instances of 'difficulty in Parsing' discussed in this book. The first case has to do with the so-called major subject construction and the second case with the OS (Object Subject) construction.
2.2.2.2.1.The major subject construction
Consider the schematic form of a surface string in (13).
(13)A-ga B-o V
'A Verb B'
Given the possibility of the major subject construction in Japanese, such as (14) and, given the fact that covert arguments are apparently possible in Japanese, as in (15) – where ec (empty category) indicates that what corresponds to it is missing in the overt string –, it should be possible for (13) to be analyzed as corresponding to (16a) as well as to (16b).
(14)a.John-ga musuko-ga sinda.
John-nom son-nom died
'It is true of John (and not of the others) that the/his son died.'
b.John-ga aitu-ga zibun-de odotta.
John-nom that:guy-nom self-by danced
'It was true of John (and not of the others) that that guy danced himself.'
(15)a.ec tuita.
'it/he/she/they/I/we/you arrived'
b.ec ec tukutta.
'he/she/they/I/we made it/them'
(16)a.A-ga [ec(-ga)[11] B-o V]
b.A-ga B-o V
Different numerations result in different output representations. The resulting structure thus may or may not contain the empty category that would end up being the thematic subject as in (16a). Suppose that A and Bcannot stand in the structural relation of what we might call 'co-arguments' (such as the relation holding between the thematic subject and the thematic object of the same verb – John and Mary in John hit Mary, for example) in order for a particular interpretation to obtain involving A and B. If the informant parsed the sentence so as to 'go to' the numeration without the empty category, i.e., if the informant parsed the sentence as corresponding to (16b), A and B would in that case stand in the structural relation of 'co-arguments', making it impossible for SR() compatible with to obtain. The informant would in that case find the sentence unacceptable under the particular interpretation . If the informant parsed the sentence as corresponding to (16a), on the other hand, A and B would not stand in the structural relation of 'co-arguments' since the thematic subject is the ec rather than A-ga, which is the major subject. Since the necessary condition for (i.e., the requirement that A and Bdo not stand in the structural relation of 'co-arguments') is thus satisfied in that case, it would be possible for SR() compatible with to obtain; hence it would be possible for the informant to accept the sentence under .[12]
It seems that, being presented a sentence of the form in (13), some informants can 'analyze it' as an instance of the major subject construction while others cannot, at least until they have been made aware of the possibility, resulting in different judgments of the sort just alluded to. An actual instance of the structural ambiguity of this sort, the parsing difficulty associated with it, and the different judgments by the informants as the result will be discussed in chapter 4: xx, in relation to the so-called Principle B effects in Japanese.
2.2.2.2.2.The OS (Object Subject) construction
Now consider the schematic form of a surface string in (17).
(17)[ ... B ... ]-o A-ga V
The 'structural ambiguity' of the OS (Object Subject) construction, such as (17), has been discussed extensively since the mid 1980s. What has emerged is the recognition that the surface string of the form in (17) can correspond to the LF representation schematized in (18), in which the o-marked phrase containing B is in the canonical thematic object position at LF.
(18)A-ga [ ... B ... ]-o V
We assume that the thematic subject position asymmetrically c-commands the thematic object position.[13] It is, however, also possible, that (17) corresponds to the LF representation in which the o-marked phrase is not c-commanded by A. [According to the analysis of the OS construction in Japanese adopted here (Ueyama 1998, 2003), the two different LF representations alluded to above obtain from different numerations and the LF representation in which the o-marked NP is not c-commanded by A results from the numeration that includes the empty category; cf. chapter xx: xx for more details.<= You might want to be a little more concrete here, giving some schematic representations like (17) and (18). I can understand what you mean quite easily because I know the Ueyama analysis, but I wonder if the general audience might not have some difficulty.] The choice between the two LF representations is thus a consequence of the choice of the numeration and hence is expected to be affected by parsing.[14]
Suppose that in order for interpretation involving A and B in (17) to be possible, the numeration without the empty category must be chosen. If the informant tends to 'go to' a numeration with the empty category for the OS construction, SR() compatible with is unlikely to obtain for the informant. The informant in that case would likely find the sentence under the particular interpretation unacceptable. On the other hand, if the informant parses the OS construction in (17) so as to 'go to' the numeration without the empty category, SR() compatible with can obtain. And in this case, it would be possible for the informant to accept the sentence under .[15] Chapter 2: xx contains discussion of concrete empirical materials pertaining to the issue just noted, in relation to the so-called reconstruction effects of BVA (bound variable anaphora) in the OS construction, which will be addressed again in chapter 4 in light of the methodological proposals to be made in this chapter.
We have just addressed cases where the informant may fail to parse sentence so as to obtain SR() compatible with interpretation . That would necessarily result in the informant finding with interpretation unacceptable. It is, however, possible that the informant tries a different parsing possibility and comes up with the 'right numeration' for obtaining the 'intended SR'. If that happens, the informant might well report that sentence is acceptable under interpretation . Depending upon the degree of difficulty in 'going to' the 'right numeration', the informants may report that sentence is not fully acceptable or is only marginally acceptable under interpretation .[16]
2.2.2.3.Difficulty caused by the unnaturalness of the entire SR()
Let us turn to (12b), repeated below
(12)b.We cannot rule out the possibility that the informant's reported judgment on the acceptability of sentence under interpretation has been unaffected by the possible unnaturalness of the entire SR().
The unnaturalness of the entire SR()may be independent of [interpretation You might want to remind the reader that this is an abbreviation of “part of the interpretation of sentence ." For a moment I was confused about that.] itself, hence independent of the whether SR() is compatible with interpretation . For example, some informants might find that book despised itself unnatural because they find unnatural a situation in which a book despises something or other; the unnaturalness of the entire sentence in that case is independent of whether or not that book and itself can be understood to refer to the same entity – after all, the same informants would likely find that book despised this paper unnatural as well. Efforts thus have to be made to ensure as much as possible that the informant judgments are not affected by difficulty caused by the possible unnaturalness of the entire SR().